RECTANGLE-BASED APPROACHES FOR REPRESENTING FLOOD DATA IN SPATIAL INDICES

Abstract

Appropriate representation of flood data in indices is crucial for efficient future querying. A common approach is to use minimum bounding rectangles to build R-tree indices rather than storing real complex objects; however, the results may be imprecise. Herein, to improve the performance of this approach, we discuss methods for producing corresponding rectangles. The first method involves generating a set of small fixed-sized squares, and the second one attempts to initially obtain a large rectangle. The third approach recursively partitions the space into four quadrants until the given size constraint is satisfied. Experimental results based on the real flood data of two representative cities demonstrate that the quadruple partitioning method can best approximate the original area using fewer cells and is thus the recommended approach.